ZX Spectrum toastrack screen jumping fix.

[1024MAK]

When I set the frequency at pin 6 to the stated crystal frequency (17.7345 MHz), I still get the jumpin but much less, but this stops if I go above 17.38. However the RF colour goes at or above this point.

Don't forget, the ULA divides down the 17.7345 MHz by 4 to get the 4.43361875 MHz colour sub carrier frequency used by the composite video and the RF/UHF outputs. Ideally the frequency of this oscillator should be close to 17.734475 MHz. And it should be rock steady.

This is the signal I get at pin 4 of the RGB.IMG_20250325_001711_1.jpg.

That's the line / horizontal sync over multiple lines.

The positive pulses keep going low many times a second.( Seen as a rapid line flickering and moving at the 0v baseline.
I captured one below:IMG_20250325_002043_1.jpg

That's the field/vertical sync (the longer duration low level) in-between line / horizontal syncs.

The frequency in the low states does not appear to change with the increase or cessation of the screen jumping when I increase or decrease the crystal frequency, but I can't tell if they are longer in time.

You need to change the timebase and trigger settings (if you can, set the trigger to only "fire" when the low pulse is bit longer than the duration of a line / horizontal sync) to almost fill the screen with the field/vertical sync and then if you 'scope has them, use the cursors or other functions to measure the length of the pulse. If the measurements are consistent even if the TV/monitor is not happy, that points more to the TV/monitor not being happy with the signal rather than the ZX Spectrum having a fault.

Most TVs/monitors are tolerant of non-standard video signals such as that from a ZX Spectrum. But not all are. The thing is, the ZX Spectrum only produces one video field, then repeats the same field. Normally with PAL video, an "odd" video field is followed by an "even" video field (interlaced) to produce one frame. These "odd" and "even" video fields have slightly different number of video lines (including half lines) so they don't overlap vertically and hence have slightly different timing compared to the video fields from a ZX Spectrum. This is why I suggested trying a different TV/monitor.

Mark

[1024MAK]

ZX Spectrum 48K Video Waveforms
These (now rather old) 'scope shots of mine may help you understand video waveforms. Keep in mind that some of them are inverted (as that's some of the signals in a 48K machine).

Mark

[Skeet]

Thank you v much, Mark. A lot to learn there....
I just tried it on the CRT and there is no jumping at any oscillator frequency set. Still goes black and white when too far off. This TV only has RF or composite inputs though, so can't test the RGB signal.
It could be the TV I'm using, but it is fine with several other zx non toastrack spectrums that I have.
Also strange that I can stop it jumping on RGB by setting the frequency above 17.7345, say 17.370, but then the colour goes from the RF signal.
I found some cheap 74S04s on eBay, so I'll try changing that
I'll composite mod it as well to see if that sorts the problem out. The RF video quality is quite bleedy anyway.
I'll digest your helpful advice and try those measurements tomorrow.
Cheers

[1024MAK]

Be careful when buying semiconductors on eBay or other trading sites.

In the Far East it’s becoming increasingly common for cheap items to be remarked or fake. So a lower speed Z80 may be cleaned up, the existing markings removed, then new markings applied with a newer, faster part number.

The same applies to other chips including some 74 series logic chips.

The reason that a 74S04 is being used is because this is a fast 74 series logic chip. A far more common (slower) 74LS04 will not work correctly in this position for example.

If the output signal from the 74S04 is stable, is the correct frequency and is consistent, then I don’t think there is anything wrong with it.

Note that the rubber key models and the ZX Spectrum+ all use a different method for generating the colour sub carrier frequency. They use an entirely separate oscillator, not the master oscillator that is used for the pixel and timing clock.

Mark

[Skeet]

Good point re fake ICs, Mark.
I have been scammed before, once with disasterous results.
The ICs are from a UK reputable seller. I've never had any problems with them before. I have a logic IC tester and they accept returns.
So no harm in trying a new chip.
Very interesting about the different generation of the colour sub carrier frequency on the 128k compared to the 48k spectrums. I see the 128k only has one crystal and the 48k has two.
Maybe it is my benchTV just struggling with the 128k video signal.
It will be interesting to see if the RGB and composite video are both stable at the same oscillator frequency when I mod it. I know most people like to keep the toastrack "original", but composite video is much easier to live with and better video quality.